Optimal Storage Regulation in a Multiple-reservoir System using Deterministic DDDP

Abstract

Dynamic Programming (DP) is one of the most powerful methods to decide optimal policy in reservoir management.A variety of algorithms for its computer implementation have been presented by which numerical difficulties encountered in the direct application of DP can beeliminated.Discrete Differential Dynamic Programming (DDDP) is an iterative algorithm in which the recurrent equation of DP is used to search for an improved trajectory in the neighborhood of trial trajectory called ‘corridor’.DDDP has an inherent drawback that sometimes it may produce non-optimal solution when corridor size remains invariable during the iteration. In this paper, a methodology in that the corridor size is reduced at next step after no improvement is achieved proposed to augment possibility of attaining to optimal solution.Using the deterministic DDDP so modified, an optimization model applicable to multiple-reservoir system with the reservoirs connected in series by linkage canals is formulated by which reservoir storages are so optimally regulated that the gap between targeted and actual storages, integrated over both all the reservoirs involved and the whole period of reservoir operation, may be minimized.A demonstrative operation of the model is carried out with its application to the system containing five reservoirs. The results indicate that with successful computational sequence it is amply possible to obtain a guide to regulating the imbalances of storage by use of the model developed.